Recent work has demonstrated the viability of an AI-based pipeline for converting paper-based nuclear operating procedures into structured digital representations suitable for automated operations. Multiple pipeline approaches were evaluated, with human-in-the-loop prompt refinement producing outputs most consistent with the source documents. These results highlight that accurate conversion alone is insufficient; the structure of the resulting data fundamentally determines its usability for downstream automation.

To enable automated and semi-automated nuclear operations, procedural content must be encoded as discrete, machine-interpretable objects while preserving sequencing, dependencies, and operational intent. However, different data structures capture these relationships with varying effectiveness, directly impacting both the fidelity of the conversion process and the ability to support automated execution, decision support, and integration with plant systems.

The conversion process is inherently iterative and requires significant human judgment. The chosen representation influences analyst performance by shaping how complex procedural logic—such as branching conditions, parallel actions, and contingencies—is interpreted and encoded. No single structure is sufficient to support both efficient conversion and operational deployment.

This paper presents a foundational data structure for digital nuclear procedures and the rationale for its relational design, emphasizing its role in enabling reliable conversion and supporting future automated nuclear operations.